Perspectives on Safety
Computerized Provider Order Entry and Patient Safety
Annual Perspective 2015
Computerized provider order entry, or CPOE, is a cornerstone of patient safety efforts, and the increasingly widespread implementation of electronic health records has made it a standard practice in health care. In 2015, the literature on CPOE and safety evolved; current research reflects not only the experience of early adopters but also that of multiple health care settings and organizations. The research on CPOE makes clear that the technology reduces prescribing errors. To date, it is less certain whether CPOE reduces clinically significant adverse drug events, and alert fatigue is a growing concern. This Annual Perspective summarizes novel findings and research directions in CPOE in 2015.
Benefits and Limitations of CPOE
Data from diverse settings demonstrates clear but circumscribed benefits from CPOE. For example, in oncology, a series of studies found that CPOE, in and of itself, may not be sufficient to ensure safe chemotherapy use. One oncology study demonstrated fewer chemotherapy medication errors using CPOE, while another found no change in chemotherapy prescribing errors after CPOE implementation. On an inpatient oncology unit, specialist pharmacist review of chemotherapy orders uncovered medication errors even when CPOE was in place. A longitudinal analysis of 5 years of CPOE use in patients receiving chemotherapy noted a significant decrease, but not elimination, of medication errors. Taken together, these studies argue that CPOE implementation does not obviate the need for ongoing oversight from health care professionals with specific expertise, such as oncology pharmacists.
Finally, evidence from a multicenter outpatient study did not find a change in the rate of preventable adverse drug events following CPOE implementation. This may be because electronic health records and the ordering functions within them are often not optimized for outpatient clinician workflow or prescribing. However, it is clear that CPOE is an improvement over paper-based ordering. One study reported significant prescribing errors when physicians reverted to paper orders during a technical failure of the electronic health record. Another study offered evidence that CPOE can reduce prescribing errors for complex treatments such as parenteral nutrition.
Drawbacks of CPOE in Practice
An important contribution to the literature in 2015 was enhanced understanding of how CPOE systems work in actual practice rather than in controlled research settings. The year's most significant advance was a combined observational and simulation study. Schiff and colleagues examined thousands of medication errors involving CPOE reported to the commercial MEDMARX database, which tracks medication safety problems. The authors developed a taxonomy for the types of safety vulnerabilities caused or missed by CPOE systems and then used these data to create scenarios in which these vulnerabilities were present. Investigators then observed physicians entering these orders into various CPOE systems in a simulation environment. The observed physicians were volunteers from institutions representing 11 different CPOE systems, including prominent commercial systems as well as homegrown ones.
Trained observers documented the ease with which clinicians could enter these problematic orders and the extent to which the CPOE system generated warnings or prevented the orders. Their observations revealed that erroneous orders could easily be entered and were often undetected in commercially available electronic health records' CPOE functions. For example, one of the cases consisted of asking clinicians to enter an order for "Insulin Aspart, 60 mL SQ BID," with the units of "mL" representing a deliberate error (insulin is dosed in units, not milliliters). In 80% of observations, users were able to enter this order—usually with only minor workarounds. In 95% of observations, the CPOE systems failed to generate a warning.
These results illustrate that currently available commercial CPOE systems are not yet optimized to promote medication safety and should serve as a wake-up call to stimulate improvements. Moreover, the methods of this study can serve as a blueprint of how vendors should test CPOE systems, both prior to their implementation and in an ongoing fashion, to ensure that high-priority safety vulnerabilities are addressed. These results are consistent with the 2014 Leapfrog Hospital Survey, released this year, which also found widespread safety gaps in CPOE systems.
In 2015, the FDA released a white paper on the safety of CPOE systems. The authors summarized their results in a commentary. Their work uncovered significant usability problems with every system tested, across a wide array of functions. Among the problems described were:
- Inconsistent medication naming within and across systems;
- Poor medication search functions;
- Difficulty interpreting displays;
- Vulnerability to wrong-patient errors when multiple records were open;
- A lack of standardized alerts, an abundance of irrelevant alerts, and a lack of reasons documented for alert overrides among clinical decision support functions; and
- Medication reconciliation modules that lacked standard terms and did not easily accommodate team-based reconciliation workflows.
Beyond the challenges of individual systems, the report also identified safety concerns related to poor interoperability among electronic health records and between electronic health records and pharmacy software. Most importantly, a lack of efficient two-way communication between physician offices and pharmacies was cited as a safety concern. Finally, the authors noted the failure to systematically identify, track, and learn from CPOE errors as a major barrier to safety and improvement.
Addressing Limitations of CPOE Implementation
There is evidence that health systems are addressing some CPOE shortcomings through customized approaches. One study demonstrated that custom alerts could prevent medication-timing errors, and another showed that tailored alerts for patients 65 and older led to decreased use of high-risk medications when safer alternatives were available. This literature indicates that customization is a promising but resource-intensive approach to enhancing the safety of CPOE.
Implementation of electronic health records with CPOE remains a challenge. A review article found that both shortcomings of the technologies, as discussed above, as well as local implementation factors such as lack of training and inefficient processes, serve as barriers to realizing the potential benefits of CPOE. Cost is also a barrier to CPOE implementation, with implementation costs (including software, consultants, and internal staff costs) exceeding $1 billion for some large health systems. A cost-utility analysis published this year suggested that CPOE implementation could generate more than $130 billion in cost savings across the United States. However, the underlying assumption that preventable adverse drug events would be reduced has not been consistently borne out in the literature, and the authors assumed lower implementation costs than recent reports from health systems would indicate.
Future directions in CPOE research should include more data on usability and errors from real-world settings, including outpatient health care settings. The field needs more research on clinical decision support to help point the way to clinical decision support that is efficient and avoids alert fatigue while identifying and preventing unsafe medication prescribing. As CPOE becomes a universally employed technology, pinpointing system features and implementation factors that promote patient safety will be critical to achieving its long-promised, and eagerly awaited, benefits.